.png){kind=icon}
Return
Zero-shot diverse audio captioning with diffusion models
.png){kind=icon}
PRE
AI
DOI：10.1016/j.knosys.2025.115205.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
• Combine diversity and zero-shot (ZS) abilities in audio captioning for the first time. • Discover noise-induced condition degradation and tackle it with condition tiling. • Adapt alignment models without audio to enable retrieval-guided Langevin dynamics. • On major benchmarks, boost diversity over prior ZS work and fix baseline accuracy gap.
Journal
K
IF:
7.6
Papers:
1.3W
Citations:
4.5W
Organization
Cited Papers
Diff-PC: Identity-preserving and 3D-aware controllable diffusion for zero-shot portrait customization.png)
INFORMATION FUSION
INFORMATION FUSIONIF15.5
PREAI
ShareSave
ShareSaveAuffusion: Leveraging the Power of Diffusion and Large Language Models for Text-to-Audio Generation
PREAI
ShareSaveShareSave
View more